Hafsteinn Einarsson

Associate Professor, University of Iceland
Research Scientist, deCODE genetics

I lead a research group working on natural-language processing for low-resource Germanic languages — Icelandic and Faroese — with active projects in computer vision for the natural sciences, clinical AI, and human genetics.

News

2026-04
New paper on temporal vision-language features for fish classification (Ecological Informatics).
2026-02
Microsoft Lingua grant funded: developing LLM red-teaming datasets for Icelandic.
2025-09
MazeEval preprint: a benchmark for sequential decision-making in language models.
2025-05
FoQA: A Faroese question-answering dataset (RESOURCEFUL 2025).
2025-01
Co-organized NB-REAL 2025 (Nordic-Baltic Responsible Evaluation and Alignment of LMs) at NoDaLiDa / Baltic-HLT.

Selected publications

All publications →

NLP for Icelandic & Germanic Selected

FoQA: A Faroese question-answering dataset

Simonsen, A., Nielsen, D. S., Einarsson, H. · Proceedings of the third workshop on resources and representations for under-resourced languages and domains (RESOURCEFUL-2025) · 2025

Introduces FoQA, the first dedicated question-answering dataset for Faroese, with extractive and generative variants. The dataset is constructed via translation and adaptation of existing QA benchmarks combined with native-speaker validation, and is accompanied by baseline evaluations of multilingual and Faroese- tuned language models. Establishes a reference benchmark for Faroese reading comprehension. RESOURCEFUL 2025 (NoDaLiDa co-located workshop).

PDF →

NLP for Icelandic & Germanic Selected

Application of ChatGPT for automated problem reframing across academic domains

Einarsson, H., Lund, S. H., Jónsdóttir, A. H. · Computers and Education: Artificial Intelligence · 2024

Studies the use of ChatGPT for automated problem reframing in academic settings, where the same underlying task is rewritten as a series of alternative problem statements to support student learning and assessment design. Reports on prompting strategies, quality control, and human evaluation across multiple academic domains. First-author paper in Computers and Education: Artificial Intelligence; one of the early systematic studies of generative AI in higher-education problem authoring.

DOI →

Genetics Selected

Sequence variants associated with BMI affect disease risk through BMI itself

Einarsson, G., Thorleifsson, G., Steinthorsdottir, V., Zink, F., Helgason, H., Olafsdottir, T., Rognvaldsson, S., Tragante, V., Ulfarsson, M. O., Sveinbjornsson, G., others, · Nature Communications · 2024

deCODE-led genome-wide study (Nature Communications 2024) showing that sequence variants associated with body mass index influence downstream disease risk primarily through BMI itself, rather than via independent pleiotropic effects. Uses Mendelian-randomisation and stratified-association analyses across the Icelandic population. Among the few Nature Communications papers in the collection.

DOI →